Observation of the Keyhole during
Plasma Arc Welding
Once the keyhole was established, the width of the keyhole did not change with an
increasing welding current and a decreasing welding speed
ABSTRACT. Keyhole plasma arc welding
is a unique arc welding process for deep
penetration. To ensure the quality of the
welds, the presence of the keyhole is crit-
ical. Understanding of the keyhole will
certainly benefit the improvement of the
process and weld quality. Currently, the
size of the keyhole is assumed to be cor-
relative with the robustness of the key-
hole process in maintaining the keyhole.
To verify this assumption, the keyhole
and the weld pool were simultaneously
monitored from the back side of the
workpiece. It was found that once the
keyhole is established, the width of the
keyhole does not change with an in-
creasing welding current and a decreas-
ing welding speed. This implies the width
of the keyhole gives no adequate infor-
mation about the stable state of the key-
hole and should not be used as an indi-
cation of the robustness of the keyhole
process in plasma arc welding.
Introduction
Keyhole plasma arc welding (PAW)
(Ref. 1) offers significant advantages over
conventional gas tungsten arc welding
(GTAW) in terms of penetration depth,
joint preparation and thermal distortion
(Ref. 2). Although its energy is less dense
than laser beam welding (LBW) (Ref. 3)
and electron beam welding (EBW) (Ref.
3), keyhole PAW is more cost effective
and more tolerant of joint preparation
(Ref. 4). Because of these distinct attrib-
utes, keyhole PAW has found applica-
tions on the welding of structural steels
(Ref. 5), automobiles (Ref. 6), airplanes
(Ref. 7), rockets (Ref. 8), space shuttles
(Ref. 9) and possibly on welding in space
(Ref. 10).
Although keyhole PAW had poten-
tial to replace GTAW (Ref. 1) in many
applications as a primary process for
precise joining, its complexity and
Y. M. ZHANG and S. B. ZHANG are with the
Welding Research and Development Labora-
tory, Center for Robotics and Manufacturing
Systems, University of Kentucky, Lexington,
Ky.
BY Y. M. ZHANG AND S. B. ZHANG
equipment cost slowed this expansion
initially (Ref. 1). During the late 1970s,
remarkable achievements were made
in simplifying the plasma cutting
process, which originally was just as
complex to use as PAW (Ref. 1). As a re-
sult, the complexity and equipment
cost of PAW were significantly reduced.
Now, cost and complexity are no longer
the major factors preventing the expan-
sion of PAW to further industrial appli-
cations.
The stable state of the keyhole is an
important issue in applying PAW (Ref. 5).
Although the stability of the keyhole has
been a concern in many research works
(Refs. 5, 11–16), no accurate definition
has been given. Concerning weld quality
control, the minimum requirement is the
maintenance of the existence of the key-
hole. It seems that precise control of key-
hole size could be an effective approach
to achieving a stable keyhole process and
quality welds. Thus, studies have been
conducted to correlate the keyhole size
to the welding parameters that determine
weld quality (Refs. 10, 11, 16, 17).
In a paper by Tomsic and Jackson (Ref.
11), the plasma arc was terminated dur-
ing welding. The workpiece was then
sectioned to measure the keyhole. The
results were both the back-side width
and the weld-face width of the keyhole
increased as the current increased or the
welding speed decreased (Ref. 11). This
suggested the widths of the keyhole as in-
dicators of the stable state of the keyhole
PAW process. Metcalfe and Quigley
mounted a photo-transistor at the end of
the workpiece to monitor the efflux
KEY WORDS
Bead-on-Plate Welds
Butt Joint Welds
Flow Rate
Keyhole
Stainless Steel
Orifice Diameter
Plasma Arc Welding
WELDING RESEARCH SUPPLEMENT | 53-s
RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT
plasma from the underside (Ref. 16). The
welding current was changed to establish
different modes associated with plasma
arc welding: stable keyhole, unstable
keyhole and nonkeyhole. They found
that the photoelectric measurement of
the light emitted from the underside of
the keyhole offered a basis for monitor-
ing the keyhole PAW process (Ref. 16). In
more recent efforts, keyhole PAW
processes were numerically simulated
(Refs. 4, 10, 17–20). However, only
Keanini and Rubinsky extensively stud-
ied the correlation between the keyhole
and the welding parameters during sta-
tionary welding (Ref. 10).
Due to a recent development
in imaging technology (Ref. 21), the arc
welding process can now be clearly
observed despite the arc light. It is
expected that clear observation of the
keyhole and the weld pool will provide
more reliable data for analyzing the
correlation between the geometry of
the keyhole and its stable state or the
robustness of the process in
maintaining the keyhole. The findings
would be fundamental in guiding the
development of sensing and control
technologies for keyhole PAW, thus
facilitating the applications of this
unique process.
Experimentation
Experimental Setup
The experimental setup is
shown in Fig. 1. The power supply is an
inverter designed for gas tungsten arc
welding and plasma arc welding. Its
current ranges from 10 to 200 A and is
precisely controlled by an inner-loop
controller. The host computer adjusts
the welding current through the analog
output interface to the power supply.
The torch, a regular commercial
straight-polarity plasma arc welding
torch rated at 200 A, and the camera
are attached to a manipulator. The
motion of the manipulator is computer
controlled.
An ultra-high shutter speed vision sys-

RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT
Fig. 1— Experimental Setup.
tem (Ref. 21) was used to simultaneously
image the keyhole and the weld pool
from the back side of the workpiece —
Fig. 1. The camera system consisted of a
strobe-illumination unit (pulse laser),
camera head and system controller. The
pulse duration of the laser was 3 ns, and
the camera was synchronized with the
laser pulse. Thus, the intensity of laser il-
lumination during the peak was much
higher than that of the plasma. Using this
vision system, the weld pool would al-
ways be clearly observed and the plasma
from the keyhole completely eliminated
from the image (Ref. 22). In this study,
both the keyhole and the weld pool need
to be imaged. By increasing the illumi-
nation area of the laser, the brightness of
the laser illumination could be reduced
so that it was close to the brightness of the
plasma. Both the keyhole and the weld
pool could, therefore, be imaged clearly
and simultaneously — Fig. 2. The weld-
face width of the weld pool was mea-
sured off-line after the experiment using
a structured-light 3-D vision sensor and a
vision algorithm developed in previous
work (Ref. 23).
Experimental Procedure
Bead-on-plate and butt-joint welds
were made on 3-mm-thick stainless steel
(304) plates in the flat position. Pure
argon was used as the shielding gas and
the plasma gas. The back side of the work-
piece was also shielded using pure argon.
54-s | FEBRUARY 1999
Keyhole
Weld pool
Oxidized
weld
Fig. 2 — Simultaneous imaging of weld pool
and keyhole.
The flow rate of the
shielding gas was 16.5
L/min (35 ft3/h) and 9.4
L/min (20 ft3/h) for the
weld face and the back
side, respectively. The
flow rate of the plasma
gas, the welding cur-
rent and the welding
speed ranged from 1.3
L/min (2.8 ft3/h) to 2.6 Fig. 3 — Geometrical parameters of weld pool and keyhole.
L/min (5.5 ft3/h), from
55 to 95 A and from 1
to 3.5 mm/s. withstanding the perturbations in welding
parameters and conditions for maintain-
Results and Discussion ing a stable keyhole.
The geometrical parameters defined in
Assume the keyhole has been estab- Fig. 3 will be used to illustrate the exper-
lished under certain welding parameters imental results. They include the weld-
and conditions. If we decrease the weld- face width of the weld pool (w), the back-
ing current or increase the welding speed, side width of the weld pool (wb) and the
the keyhole will eventually collapse. Or, back-side width of the keyhole (wh).
if we increase the welding current or de-
crease the weld speed, burn-through (a Welding Current and Speed
hole in the root bead) may occur. Of
course, the changes in other welding pa- To verify the suggestion that the size
rameters or conditions may also cause the of the keyhole is correlative with the sta-
collapse and/or burn-through. The mini- ble state of the keyhole process, experi-
mum changes in the welding parameters ments were conducted using varying
or conditions that result in either collapse welding parameters. The method was to
or burn-through can quantify the capabil- change the stable state of the keyhole and
ity or the robustness of the process in then examine whether there is a corre-
maintaining a stable keyhole or the stable sponding change in the keyhole size. The
state of the keyhole process. In the fol- keyhole size was measured by using the
lowing discussion, the stable state of the back-side width of the keyhole as defined
keyhole process will be used as a term to in Fig. 3. The welding speed and welding
describe the robustness of the process current were changed to alter the stable
 A B

RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT
C D

Fig. 4 — Weld pool and keyhole under varying welding parameters. Plasma gas flow: 1.3 L/min (2.8 ft3/h), orifice diameter: 1.57 mm (0.062 in.). A
— Bead on plate, welding speed 2 mm/s; B — bead on plate, welding speed 2.8 mm/s; C— bead on plate, welding current 65 A; D — butt-joint
weld, welding speed 2.2 mm/s.

state of the keyhole process. through or a collapse is about to

Figure 4 shows the results of four ex-
periments conducted using an increasing
current or a decreasing speed. Both
bead-on-plate and butt-joint welds were
made. The flow rate of the plasma gas
and the diameter of the orifice were the
same for the four experiments. As shown
in Fig. 4, when the welding current in-
creased or the welding speed decreased
to certain levels, the keyhole was estab-
lished. Although the further increase in
the welding current or the further de-
crease in the welding speed increased w
and wb, the width of the keyhole (wh) did
not change correspondingly: wh re-
mained at approximately 1 mm. The
same phenomenon can be observed
from the images given in Fig. 5.
The experimental results in Figs. 4 and
5 showed that the width of the keyhole was
not changed by varying the welding cur-
rent or the welding speed when no burn-
through or collapse occurred. It would be
interesting to examine whether the width
of the keyhole will change if a burn-
occur. A B
In Fig. 6, the burn-through
eventually takes place for all four
experiments. The cause of the
burn-through was the increase in
the welding current or the de-
crease in the welding speed, and
both bead-on-plate and butt joint
weld experiments were con-
ducted. Experimental results
C D
clearly demonstrated that the
width of the keyhole did not pro-
vide any information or indication
about the burn-through that was
about to occur. Also, experiments
have been performed using a de-
creasing welding current or an in-
creasing welding speed to cause a
collapse of the keyhole. An exam-
ple is illustrated in Fig. 7. It can be
Fig. 5 — Back-side view of keyhole and weld pool at dif-
seen that the width of the keyhole ferent currents and welding speeds. Plasma gas flow: 1.3
failed to predict the collapse. L/min (2.8 ft3/h), orifice diameter: 1.57 mm (0.062 in.).
When the welding current in- A — Welding speed 2 mm/s, welding current 60 A; B—
creases or the welding speed de- welding speed 2 mm/s, welding current 65 A; C — weld-
ing speed 1.8 mm/s, welding current 70 A; D — weld-
creases, the stable state of the key- ing speed 1.6 mm/s, welding current 75 A.

WELDING RESEARCH SUPPLEMENT | 55-s

 A B
RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT

C
D

Fig. 6 — Burn-through experiments. Orifice diameter: 1.57 mm (0.062 in). A — Bead on plate, welding current 55 A, plasma gas flow 1.3 L/min (2.8
ft3/h); B — butt-joint weld, welding current 63 A, plasma gas-flow rate 1.6 L/min (3.4 ft3/h); C — bead on plate, welding speed 2 mm/s, plasma gas-
flow rate 1.3 L/min (2.8 ft3/h); D— butt-joint weld, welding speed 1.6 mm/s, plasma gas-flow rate 1.3 L/min (2.8 ft3/h).

Fig. 7 — Keyhole collapse

due to decrease in welding
current. Butt-joint weld,
welding speed: 2.4 mm/s,
plasma gas-flow rate: 1.3
L/min (2.8 ft3/h), orifice di-
ameter: 1.57 mm (0.062 in).

Fig. 8 — Horizontal forces acting on keyhole

wall.

56-s | FEBRUARY 1999

 A B
Fig. 9 — Plasma keyhole arc welding using varied rate of plasma gas flow (bead on plate). A — Welding speed 2.8 mm/s, welding current 65 A,
orifice diameter 1.57 mm (0.062 in); B — welding speed 3 mm/s, welding current 95 A, orifice diameter 2.36 mm (0.093 in).
hole changes accordingly. For example,
when the keyhole is barely established, a
slight decrease in the welding current
and/or the flow rate of the plasma gas or
a slight increase in the welding speed may
result in a collapse of the keyhole. This
implies that the stable state of the keyhole
needs to be improved to prevent the col-
lapse. If the welding current and the flow
rate of the plasma gas are increased
and/or the welding speed is decreased,
the process will withstand larger varia-
tions in the welding parameters, such as
the welding current, the flow rate of the
plasma gas and the welding speed, with-
out a collapse of the keyhole. The stable
state of the keyhole is improved. After the
welding current and the flow rate of the
plasma gas increase and/or the welding
speed decreases to certain levels, further
increase in the welding current and the
flow rate of the plasma gas or decrease in
the welding speed may cause a burn-
through. The stable state degrades.
Hence, when the welding current or the
welding speed changes, the stable state of
the keyhole process in terms of maintain-
ing a stable keyhole changes accordingly.
Experimental results in Figs. 4, 6 and
7 demonstrated, when the stable state of
the keyhole varied due to a change in the
welding current or in the welding speed,
the width of the keyhole remained con-
stant. Therefore, the width of the keyhole
is not correlative with the stable state of
the keyhole process or the robustness of
the process in maintaining a stable key-
hole. Also, full penetration is measured
using the back-side bead width of the
weld pool (Refs. 24, 25). In the experi-
ments, the back-side bead width wb did
change when the welding parameters
were changed. Therefore, in addition to
the stable state of the keyhole process,
the width of the keyhole is also not cor-
relative with the weld penetration.
Analysis
The independence of the width of the
keyhole on the welding current and the
welding speed is quite understandable.
In fact, once the keyhole is established,
the plasma gas passes through the key-
hole. Because the majority of the weld-
ing current earths at the top surface of the
weld pool (Ref. 26), the stagnation pres-
sure of the plasma gas flow plays a major
role in balancing with the surface tension
pressure plus the hydrostatic head to
keep the keyhole open (Refs. 10, 16).
The horizontal forces acting on the
keyhole wall are shown in Fig. 8. The sur-
face tension pressure that tends to close
the keyhole is (Refs. 16, 27)
Ps=Ts/r (1)
where Ts is the surface tension constant
and r is the radius of the keyhole. (The
keyhole is assumed to be axially sym-
metric.) The pressure generated by the
hydrostatic head of the melted weld
metal is (Ref. 16)
Pm = ρmgz (2)
where ρm is the density of the liquid
metal, g is the gravity acceleration and z
is the coordinate along the thickness di-
rection — Fig. 8. At the bottom surface of
the plate, r = wh/2 and z=h. Divide the
keyhole as thin circular elements along
WELDING RESEARCH SUPPLEMENT | 57-s
RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT
the z direction. The thickness of a thin
circular element is ∆z. Assume that the
mass of the plasma gas in any ∆z, de-
noted by ∆m, does not change with z. It
is further assumed that ∆m/∆z is only de-
termined by the flow rate of the plasma
gas fp and the speed of the plasma gas
flow u. When fp is given, u depends on
the diameter of the orifice. From the ideal
gas law (Ref. 28), the plasma gas stagna-
tion pressure Pg can be expressed as
(3)
where ṁ is defined as∆m/∆z, and k’ and
K are coefficients. When the keyhole is
stationary, the stagnation pressure of the
plasma gas flow should balance the sur-
face tension pressure and the pressure
generated by the hydrostatic head. Equa-
tions 1–3 show that for the given mater-
ial and plate thickness h, the back-side
width of the keyhole is only controlled by
Ts, fp and u.
The surface tension constant Ts is tem-
perature dependent. When the welding
current or the welding speed changes,
the temperature of the plasma, and,
therefore, the temperature of the keyhole
surface, should change. However, for a
given material, one may assume that the
temperature of the keyhole surface, es-
pecially at the bottom of the keyhole, is
not subjected to a severe change. Hence,
once the keyhole is established and the
heat input is sufficient to melt enough
material, changes in both the welding
current and the welding speed should not
 have a substantial influence on the back-
side width of the keyhole.
That the width of the keyhole is not
dependent on the welding current has
RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT/RESEARCH/DEVELOPMENT
also been indirectly obtained for station-
ary keyhole PAW by Keanini and Rubin-
sky (Ref. 11) based on numerical analy-
sis. In their study, the heat input was
measured by the initial plasma tempera-
ture rather than the welding current and
the welding speed. It was found that the
diameter of the keyhole, thus, the width,
at any given axial position is independent
of the initial plasma temperature (Ref.
11). Although the study did not attempt
to correlate the initial plasma tempera-
ture with the electric power inputs, they
claimed that plasma isotherms for vari-
ous arc power inputs nevertheless con-
firmed that the initial plasma conditions
are comparable to the initial plasma tem-
perature. Keanini and Rubinsky’s numer-
ical analysis supports the fact that the
back-side width of the keyhole is inde-
pendent of the welding current during
stationary keyhole PAW.
Flow Rate and Orifice
Although the back-side width of the
keyhole is independent of the welding
current and the welding speed, the flow
rate of the plasma gas and the diameter of
the orifice both have significant influences
on the width of the keyhole. This can be
seen from the dependence of the stagna-
tion pressure of the plasma gas on the flow
speed, which can be changed by altering
the diameter of the orifice and the flow
rate of the plasma gas. As can be observed
from Fig. 9, for a given orifice, the back-
side width of the keyhole increases as the
flow rate of the plasma gas increases. Also,
a larger width of the keyhole resulted in
Fig. 9B due to the larger diameter of the
orifice. It is known that the stable state of
the keyhole changes with the flow rate of
the plasma gas and the diameter of the ori-
fice. Thus, in the case where a change in
the stable state is caused by the plasma gas
or the diameter of the orifice, the width of
the keyhole gives information about the
change in the stable state. However, the
stable state of the keyhole process may be
caused by different welding parameters,
including the welding current and the
welding speed. Therefore, in general, the
width of the keyhole and the stable state
of the keyhole process are not inherently
correlated.
Summary
Bead-on-plate and butt-joint weld ex-
periments have been conducted on 3-
mm-thick stainless steel plates to verify
58-s | FEBRUARY 1999
the feasibility of using the back-side
width of the keyhole as a measurement
of the stable state of the keyhole process.
Experimental results revealed that once
the keyhole is established, the width of
the keyhole does not change with the
changes in the welding current and the
welding speed, but it does change with
the changes in the flow rate of the plasma
gas and the diameter of the orifice. On
the other hand, the stable state of the key-
hole process and the weld-joint penetra-
tion vary when either the welding cur-
rent, the welding speed or the flow rate
of the plasma gas changes. Hence, the
width of the keyhole gives no adequate
information on the state of the keyhole
process and the weld-joint penetration.
Also, it provides no predictions of the
burn-through and the collapse that are
about to occur. Therefore, in general, the
width of the keyhole should not be rec-
ommended as a critical parameter to
monitor and control either the keyhole
process or the weld-joint penetration.
References
1. Craig, E. 1988. The plasma arc welding
— a review. Welding Journal 67(2): 19–25.
2. Tomsic, M., and Barhorst, S. 1984. Key-
hole plasma arc welding of aluminum with
variable polarity power. Welding Journal
63(2): 25–32.
3. Welding Handbook, Vol. 2: Welding
Processes. 1991. 8th edition, American Weld-
ing Society, Miami, Fla.
4. Hsu, Y. F., and Rubinsky, B. 1988. Two-
dimensional heat transfer study on the keyhole
plasma arc welding process. International
Journal of Heat and Mass Transfer 31(7):
1409–1421.
5. Martikainen, J. K., and Moisio, T. J. I.
1993. Investigation of the effect of welding pa-
rameters on weld quality of plasma arc key-
hole welding of structural steels. Welding
Journal 72(7): 329-s to 340-s.
6. Vilkas, E. P. 1991. Plasma arc welding of
exhaust pipe system components. Welding
Journal 70(4): 49–52.
7. Irving, B. 1997. Why aren’t airplanes
welded? Welding Journal 76(1): 31–41.
8. Irving, B. 1992. Plasma arc welding
takes on the advanced solid rocket motor.
Welding Journal 71(12): 49–50.
9. Nunes, A. C., et al. 1984. Variable po-
larity plasma arc welding on the space shuttle
external tank. Welding Journal 63(9): 27–35.
10. Keanini, R. G., and Rubinsky, B. 1990.
Plasma arc welding under normal and zero
gravity. Welding Journal 69(6): 41–50.
11. Tomsic, M. J., and Jackson, C. E. 1974.
Energy distribution in keyhole mode plasma arc
welds. Welding Journal 53(3): 109-s to 115-s.
12. Bashenko, V. V., and Sosnin, N. A.
1988. Optimization of the plasma arc welding
process. Welding Journal 67(10): 233-s to
237-s.
13. Martinez, L. F., et al. 1992. Front side
keyhole detection in aluminum alloys. Weld-
ing Journal 71(5): 49–52.
14. Martikainen, J. 1995. Conditions for
achieving high-quality welds in the plasma-
arc keyhole welding of structural steels. Jour-
nal of Materials Processing Technology 52:
68–75.
15. Halmoy, E., Fostervoll, H., and Rams-
land, A. R. 1994. New applications of plasma
keyhole welding. Welding in the World 34:
285–291.
16. Metcalfe, J. C., and Quigley, M. B. C.
1975. Keyhole stability in plasma arc welding.
Welding Journal 54(11): 401-s to 404-s.
17. Kim, C.-J. 1994. Parametric study of
the two-dimensional keyhole model for high
power density welding processes. ASME Jour-
nal of Heat Transfer 116(1): 209–214.
18. Dowden, J., and Kapadia, P. 1994.
Plasma arc welding: a mathematical model of
the arc. Journal of Physics (D): Applied Physics
27: 902–910.
19. Nehad, A.-K. 1995. Enthalpy tech-
nique for solution of stefan problems: appli-
cation to the keyhole plasma arc welding
process involving moving heat source. Inter-
national Communications in Heat and Mass
Transfer 22(6): 779–790.
20. Hung, R. J., and Long, Y. T. 1996. Math-
ematical model of variable polarity plasma arc
welding process. Proceedings of the National
Science Council (A): Physical Science and En-
gineering 20(1): 90–109.
21. Hoffman, T. 1991. Real-time imaging
for process control. Advanced Material &
Processes 140(3): 37–43.
22. Kovacevic, R., Zhang, Y. M., and L. Li.
1996. Monitoring of weld penetration based
on weld pool geometrical appearance. Weld-
ing Journal 75(10): 317-s to 329-s.
23. Zhang, Y. M., and Kovacevic, R. 1997.
Real-time sensing of sag geometry during GTA
welding. ASME Journal of Manufacturing Sci-
ence and Engineering 119(2): 151–160.
24. Zacksenhouse, M., and Hardt, D. E.
1983. Welding pool impedance identification
for size measurement and control. ASME Jour-
nal of Dynamic Systems, Measurement, and
Control 105(3): 179–184.
25. Zhang, Y. M., Wu, L., Walcott, B. L.,
and Chen, D. H. 1993. Determining joint pen-
etration in GTAW with vision sensing of weld-
face geometry. Welding Journal 72(10): 463-s
to 469-s.
26. Dowden, J., Kapadia, P., and Fenn, B.
1993. Space charge in plasma arc welding and
cutting. Journal of Physics (D): Applied
Physics 26: 1215–1223.
27. Kroos, J., Gratzke, U., and Simon, G.
1993. Towards a self-consistent model of the
keyhole in penetration laser beam welding.
Journal of Physics (D): Applied Physics 26:
474–480.
28. Evett, J. B., and Liu, C. 1987. Funda-
mentals of Fluid Mechanics, McGraw-Hill
Book Company, New York, N.Y.